Identifying risk for recidivism among partner violent men reported to the Swedish police

(1)

Identifying risk for recidivism

among partner violent men reported

to the Swedish police

Joakim Petersson

Main supervisor: Eija Viitasara

Co-supervisors: Susanne Strand & Heidi Selenius

Faculty of Human Sciences

Thesis for Doctoral degree in Health sciences Mid Sweden University

(2)

Akademisk avhandling som med tillstånd av Mittuniversitetet i Sundsvall framläggs till offentlig granskning för avläggande av filosofie doktorsexamen fredagen 15 maj 10.00, C306, Mittuniversitet, Sundsvall och Hörsal L2 Örebro universitet. Seminariet kommer att hållas på svenska via videolänk (zoom).

Identifying risk for recidivism among partner violent men reported to the Swedish police

Printed by Mid Sweden University, Sundsvall ISSN:1652-893X

ISBN:978-91-88947-44-4 Faculty of Human Sciences Mid Sweden University,SE-851 70 Phone: +46 (0)10 142 80 00

(3)

i

ACKNOWLEDGEMENTS

I would like to express my gratitude to my supervisors, Eija Viitasara, Susanne Strand, and Heidi Selenius for all their help and guidance during my time as a doctoral student. To this end, I would especially like to thank Susanne Strand and Heidi Selenius who have been with me since the beginning of my doctoral studies. You have both been an immense support throughout these years, providing valuable guidance, counseling, and help whenever needed.

I would also like to thank my former colleagues at the Department of Social Sciences, Mid Sweden University, and my colleagues at the Department of Health Sciences, Mid Sweden University, for their support and interest in my work. Special thanks to the colleagues who have reviewed the various draft versions of this thesis.

To my current colleagues at the School of Law, Psychology and Social Work, Örebro University, and especially within the criminology subject. Thank you for your support during my final stages of the doctoral student period. A special thanks to Anna Meehan, head of the criminology unit, for your constant support and valuable help in facilitating the completion of this thesis.

I also would like to extend my gratitude to the police officers involved in the research project in Jämtland and Västernorrland. You have been very accommodating and helpful during this eight-year long research project, and without your extremely important work and commitment to prevent intimate partner violence, this thesis would not have existed.

To my parents, Bo and Marie-Louise, and my sisters Jasmine and Josefine. Your unconditional support and encouragement have been my guiding light during these years. To my family and friends, thank you for your understanding of the time-consuming process associated with writing this thesis and for reminding me that there is more to life than work.

This thesis was made possible by a grant from the Swedish Crime Victim Compensation and Support Authority.

(4)

ii

ABSTRACT

Intimate partner violence (IPV) against women is a global public health issue, where every third woman has experienced such violence. Moreover, IPV recidivism rates are generally high. These figures indicate that the police need a better understanding of the risk factors related to those perpetrators who pose the highest risk of recidivating in IPV. To this end, research has found that IPV perpetrators who are violent towards their partner as well as others (referred to as the antisocial subtype) display more risk factors for IPV than those perpetrators who are violent only against their partner (referred to as the family-only subtype). However, there are still uncertainties whether these two subtypes differ in terms of characteristics related to recidivism (i.e., risk profile) and actual recidivism. Thus, this thesis aimed to examine differences in risk profiles and recidivism rates between the antisocial perpetrators and the family-only perpetrators.

This thesis was based on a systematic literature review and three empirical studies. The empirical studies were based on data collected from the Swedish police and consisted of IPV risk assessments. The risk assessments were performed by the police using the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER). These empirical studies relied on a sample of 657 male perpetrators who had been reported to the police and subjected to a violence risk assessment for allegedly perpetrating IPV against a female partner.

The results demonstrated that categorizing partner violent men as either antisocial or family-only can help identify the perpetrators most at risk to recidivate in IPV. As such, the antisocial perpetrators displayed a greater diversity as well as degree of risk factors for IPV, and were more likely to recidivate in IPV, despite legal interventions from the police. In contrast, the family-only perpetrators presented with fewer risk factors, were characterized as socially well-adjusted outside of the relationship, and less likely to recidivate in IPV. However, several family-only perpetrators recidivated in IPV, meaning that such perpetrators should not routinely be dismissed as low-risk perpetrators. In conclusion, the results of this thesis can be used to improve the ability of those assigned to assess risk for future IPV to identify those perpetrators most at risk to recidivate. In turn, this could enable a more informed and adequate response aiming to prevent, or at best reduce, this risk.

(5)

iii

SAMMANFATTNING

Mäns partnervåld mot kvinnor är ett globalt folkhälsoproblem där uppskattningsvis var tredje kvinna någon gång utsatts för sådant våld. Dessutom är återfall i partnervåld vanligt förekommande. Detta indikerar på att polisen behöver mer kunskap om vilka förövare som utgör den högsta risken för att återfalla i nytt partnervåld. Tidigare forskning visar på att de partnervåldsförövare som är generellt våldsamma (d.v.s. våldsamma både mot sin partner och mot andra) uppvisar fler riskfaktorer för återfall i partnervåld än de partnervåldsförövare som är våldsamma enbart mot sin partner. Dock finns det otillräcklig kunskap huruvida dessa förövare uppvisar olika riskprofiler (d.v.s. riskfaktorer och karaktäristika relaterade till återfall i partnervåld) och återfallsbenägenhet. Syftet med föreliggande avhandling var således att undersöka skillnader i riskprofiler och återfall i partnervåld mellan generellt våldsamma och icke generellt våldsamma partnervåldsförövare.

Föreliggande avhandling utgjordes dels av en systematisk litteraturöversikt, dels av tre empiriska vetenskapliga studier. De empiriska studierna baserades på riskbedömningar utförda av polisen på misstänkta partnervåldsförövare. Dessa riskbedömningar utfördes med hjälp av riskbedömningsinstrumentet Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER). Urvalet för dessa empiriska studier bestod av 657 män som anmälts till polisen för partnervåld mot en kvinna.

Resultaten påvisade att generellt våldsamma och icke generellt våldsamma partnervåldsförövare uppvisade olika riskprofiler och återfallsbenägenhet. Generellt våldsamma partnervåldsförövare uppvisade fler riskfaktorer för återfall i partnervåld och var mer benägna att återfalla i sådant våld. Icke generellt våldsamma partnervåldsförövare uppvisade färre riskfaktorer och återföll i mindre utsträckning. Dessa förövare var även mer socialt välanpassade utanför relationen. Dock fanns det flera icke generellt våldsamma partnervåldsförövare som återföll i partnervåld, vilket innebär att dessa förövare inte rutinmässigt bör betraktas som en låg-risk-grupp för återfall.

Sammanfattningsvis kan möjligheten att identifiera de förövare som löper högst risk att återfalla i nytt partnervåld förbättras om hänsyn tas till huruvida dessa förövare är generellt eller icke generellt våldsamma. Detta kan i slutändan bidra till ökade möjligheter för polisen och övriga samhället att reducera de höga återfallssiffrorna och därmed förhindra partnervåld mot kvinnor.

(6)

iv

LIST OF PAPERS

This thesis is mainly based on the following four papers, herein referred to by their Roman numerals:

Paper I Petersson, J., & Strand, S. (2018). Family-only perpetrators of intimate partner violence: A systematic review. Trauma,

Violence, & Abuse, 21(2), 367-381. doi:

10.1177/1524838018770410.

Paper II Petersson, J., Strand, S., & Selenius, H. (2019). Risk factors for intimate partner violence: A comparison of antisocial and family-only perpetrators. Journal of Interpersonal Violence, 34(2), 219-239. doi: 10.1177/0886260516640547.

Paper III Petersson, J., & Strand, S. (2017). Recidivism in intimate partner violence among antisocial and family-only perpetrators.

Criminal Justice and Behavior, 44(11), 1477-1495. doi:

10.1177/00938548177199.

Paper IV Petersson, J., & Strand, S. (2020). Characteristics and recidivism in relation to arrest: Differentiating between partner violent perpetrator subtypes. Nordic Journal of Criminology. Advance online publication. doi: 10.1080/2578983X.2020.1727119 Published papers have been reprinted with permission from the copyright holders.

(9)

vii

ABBREVIATIONS

B-SAFER Brief Spousal Assault Form for the Evaluation

of Risk

GV Generally violent

IPV Intimate partner violence

NCCP National Council for Crime Prevention

NPB National Police Board

OR Odds ratio

PO Partner only

SPJ Structured professional judgment

SRC Swedish Research Council

(10)

(11)

1

1 INTRODUCTION

Intimate partner violence (IPV) against women is a significant public health issue and a violation of human rights. According to the World Health Organization (WHO, 2013), IPV is one of the most common forms of violence experienced by women globally, and can be described as a public health problem of pandemic proportions (García-Moreno et al., 2015). Victimization of such violence is associated with a range of negative health consequences, such as physical injuries (ranging from bruises to death) and psychological trauma (e.g., stress and chronic mental health disorders: e.g., Ellsberg, Jansen, Heise, Watts, & García-Moreno, 2008; Garcia-Moreno et al., 2015). As such, women who have been victimized by IPV are more likely to use health care services than non-victimized women are.

In their duty to prevent crimes, the police are one of the primary responders to IPV and the first contact for both victims and perpetrators with the criminal justice system (Erez, 2002; Storey, Kropp, Hart, Belfrage, & Strand, 2014). Because of its high prevalence, IPV is the most common form of violence reported to the police (e.g., Belfrage & Rying, 2004). In fact, more time is spent by police officers responding to IPV-related calls than time spent investigating other crimes such as murder, rape, and assaults combined (Nietzel, 2000).

Despite the fact that the police devote a substantial amount of time and resources as well as show genuine commitment to prevent IPV from re-occurring (e.g., Strand, Petersson, Fröberg, & Storey, 2016), the recidivism rates among perpetrators of such violence are very high. As such, international studies have reported recidivism rates as high as 75% among partner violent men (e.g., Babcock, Green, & Robie, 2004; Bennett Cattaneo & Goodman, 2005). Corresponding rates in Sweden among IPV perpetrators who have been reported to the police have been found to be as high as 42% (Belfrage & Strand, 2012). These high rates clearly demonstrate that the police’s work with reducing IPV recidivism needs to be improved if they are to fulfill their duty in preventing crimes (e.g., National Council for Crime Prevention [NCCP], 2017).

A possible explanation for the inability of the police to reduce the rates of IPV recidivism could be the general tendency within the criminal justice system to view partner violent men as a homogeneous group of offenders (e.g., Sartin, Hansen, & Huss, 2006). This approach entails considering all partner violent men to share the same characteristics and reasons for using violence against a female partner. However, this view eliminates the variation in risk

(12)

2

profiles (i.e., the personal characteristics and risk factors related to risk for recidivism: Simmons & Lehmann, 2010) previously reported among partner violent men (e.g., Holtzworth-Munroe & Meehan, 2004), which could also account for variations in recidivism. Thus, the overall aim of this thesis was to examine if addressing the heterogeneity among partner violent men could help identify those perpetrators most at risk to recidivate, and thereby increase the possibility to prevent this public health problem.

1.1 Definition of IPV

The definition of IPV used in the present thesis defines such violence as “any actual, attempted, or threatened physical harm, inflicted by a man or a woman with whom the victim has or has had an intimate, sexual, relationship” (Kropp, Hart, & Belfrage, 2010, p. 2). This definition is used by the Swedish police in identifying cases of IPV (The National Police Board [NPB], 2014). Importantly, although the definition presented above is gender-neutral, implying that both men and women can be both victims and perpetrators of IPV, the majority of such perpetrators reported to the police are men (NCCP, 2019a). In fact, in the sample of police-reported IPV perpetrators used for the empirical studies in this thesis (Papers II, III, & IV), only 4% (n = 29) of perpetrators were women (Strand et al., 2016). Thus, this limited amount of female perpetrators would neither be statistically nor methodologically appropriate to analyze. Therefore, the present study focused on men's violence against a current or former female intimate partner. Moreover, due to low reporting rates, only the more severe forms of IPV are reported to the police (NCCP, 2014; Sartin et al., 2006). As such, since the current thesis is based on police-reported IPV, the type of IPV examined is by definition more severe (i.e., more severe forms of both psychological and physical IPV).

As also implied from the definition mentioned above, IPV is a multifaceted phenomenon consisting of a range of physically and non-physically violent behaviors (WHO, 2012, 2013). Physical violence consists of, but is not limited to, being slapped, pushed, having something thrown at you that could hurt you, being kicked, choked, sexually assaulted, or having a weapon used against you. Non-physically violent behaviors, in the form of threatened physical harm, is also a common form of IPV. This includes, for example, threats of violence, harassment, and stalking. Moreover, controlling and coercive behaviors used by the male partner to physically, socially, emotionally, and economically restrict and isolate a female partner can also

(13)

3

be frequently occurring behaviors in violent intimate relationships (WHO, 2012, 2013).

1.2 Prevalence of IPV

Drawing on a global systematic review of the prevalence of male-to-female IPV, the WHO (2013) estimated that approximately one-third of all ever-partnered women had ever been a victim of such violence. In the European region, about 25% of ever-partnered women reported lifetime IPV victimization. This prevalence rate corresponds to a survey on violence against women conducted within the member states of the European Union, where 22% of ever-partnered women reported lifetime experience of physical or sexual abuse by an intimate male partner (European Union Agency for Fundamental Rights, 2014).

Similar rates have been found in Sweden where the most recently conducted national survey of IPV victimization estimated that 25% of women aged 16-79 had experienced lifetime victimization of such violence (NCCP, 2014). Moreover, 8 550 cases of assault perpetrated towards a female victim by a current or former intimate male partner were reported to the Swedish police during 2019 (NCCP, 2019b). However, the national survey demonstrated that only 3.9% of those who were victimized during 2012 stated that the incident was reported to the police (NCCP, 2014) Thus, the dark figures of IPV in official crime statistics are substantial (e.g., NCCP, 2014; Sartin et al., 2006). Additionally, the WHO (2013) estimated that 38% of all murders worldwide involving a female victim was committed by a former or current intimate male partner. In Sweden, 22 women were killed by a current or former male partner during 2018 (NCCP, 2019a).

1.3 IPV recidivism

In addition to the high prevalence rates of IPV victimization, for some perpetrators violence towards a female partner is systematic and repeated. As such, recidivism rates among IPV perpetrators are generally high. For instance, in a meta-analytic review of IPV treatment intervention studies (Babcock et al., 2004), recidivism rates as high as 75% were reported. Moreover, studies examining recidivism rates in criminal justice samples have reported such rates to range between 15% and 60%. In a Spanish sample of 40 men released from prison and with an average follow-up of 21 months, 15% recidivated in IPV (Loinaz, 2014). Furthermore, following a Canadian sample

(14)

4

of 150 incarcerated men who were released back into the community, Hilton, Harris, Popham, and Lang (2010) reported that 27% recidivated in an IPV-related offense during a follow-up period of nearly eight years. Additionally, longitudinal studies of arrested male perpetrators in the US have found even higher recidivism rates. In their study of 356 male perpetrators who had been arrested for IPV, Klein and Tobin (2008) reported that 60% of the study participants were returned to court for an IPV-related offense within nine years of follow-up. Likewise, using arrest data for a sample of 286 male IPV perpetrators, Richards, Jennings, Tomsich, and Gover (2014) demonstrated that half (51.5%) of the sample recidivated in IPV over a 10-year follow-up period.

The majority of studies examining IPV recidivism rely on official records (e.g., arrest records or convictions) to measure recidivism. In general, such studies tend to underestimate the actual rates of recidivism as such crimes largely go unreported to the criminal justice system (e.g., Babcock et al., 2004; Sartin et al., 2006). Several studies have also found that men consistently underreport their use of violence when compared to victims’ self-reports, and have therefore concluded that victims’ self-report of re-victimization is a more reliable source of actual recidivism rates (e.g., Bennett Cattaneo & Goodman, 2005; Rosenfeld, 1992). In a review of recidivism in various samples, Bennett Cattaneo and Goodman (2005) found that recidivism rates in studies that relied on victim self-reports ranged between 22% and 74%. For instance, Miller and Krull (1997) found that about 74% of victims reported being re-victimized by their current or former male partner in terms of physical violence, threats, and destruction of property.

Less research regarding IPV recidivism has been carried out nationally. Albeit, the corresponding recidivism rates among male IPV perpetrators in Sweden are also high. Using official crime statistics for follow-up, studies analyzing recidivism among male IPV perpetrators reported to the police (Belfrage & Strand, 2012; Belfrage et al., 2012; Svalin, Mellgren, Torstensson Levander, & Levander, 2014, 2018) or convicted for IPV-related crimes (Grann & Wedin, 2002; Haggård, Freij, Danielsson, Wenander, & Långström, 2017) have found such rates to range between 18% and 42%. As expected, the rates of repeat victimization are even higher according to victims' self-reports. Results from the national survey demonstrated that 56% of women victimized by IPV reported experiencing abuse from their partner on repeated occasions (NCCP, 2014).

Research, both internationally and nationally, has also identified a critical period where the risk for IPV recidivism is elevated. As such, this risk is

(15)

5

heightened within the first year post-intervention, including post-treatment (e.g., Bowen, Gilchrist, & Beech, 2005; Lin et al., 2009; Petrucci, 2010), post-arrest (e.g., Richards et al., 2014), post-incarceration (e.g., Grann & Wedin, 2002; Loinaz, 2014), post-assessment with risk assessment instruments (Stanfield & Williams, 2014; Svalin et al., 2014), and post-placement on probation (Klein & Crowe, 2008). More specifically, some studies report an even shorter critical period for IPV recidivism, where the majority of perpetrators recidivate within 1 to 2 months post-intervention (Klein & Tobin, 2008; Richards et al., 2014; Svalin et al., 2014). For example, Svalin et al. (2014) reported that among those perpetrators in their sample who recidivated, 40% did so within the first month after being subjected to a structured violence risk assessment by the Swedish police.

However, although IPV recidivism rates are high, considerable variation in such recidivism rates have also been found (Babcock et al., 2004; Bennett Cattaneo & Goodman, 2005). These findings indicate that IPV for some perpetrators is a stable and systematic behavioral pattern, whereas other perpetrators are seemingly able to desist from recidivating. Unlike the commonly advocated belief that IPV always escalates in terms of severity and frequency (e.g., Cavanaugh & Gelles, 2005; Saunders, 2004), this implies that IPV perpetrators are a heterogeneous group of offenders with different violence trajectories and etiologies, which can explain their use of IPV.

1.4 Theoretical framework: A public health approach to

understanding IPV

A public health approach to understanding IPV entails an interdisciplinary approach, drawing on knowledge from a range of disciplines such as medicine, epidemiology, psychology, and criminology (WHO, 2010). Of specific relevance to the current thesis, the public health approach to understanding IPV includes drawing on a criminal justice approach to such violence. The main task of the criminal justice approach to IPV is to respond to such violence after it has occurred (i.e., secondary prevention), and relies on deterrence, incarceration, punishment and rehabilitation of IPV perpetrators to prevent and respond to such violence (WHO, 2010).

In an attempt to provide an explanation as to why and how partner violent men differ in terms of their etiologies of violence, two theories compatible with a public health approach were used. To this end, this thesis combined the integrated ecological model (Heise, 1998) commonly used in public health

(16)

6

research (WHO, 2010, 2012) with the criminological theory of stake in conformity (Toby, 1957).

1.4.1 The integrated ecological model

The notion of complexity and heterogeneity among IPV perpetrators presented above is supported by the integrated ecological model. According to the WHO (2012), this is the most widely used model in understanding IPV and has been used by several researchers to explain why such violence occur (e.g., Heise, 1998; Lauritsen & Schaum, 2004; Saunders, 2004; Silverman & Williamson, 1997; Smith Slep, Foran, Heyman, & United States Air Force Family Advocacy Research Program, 2014). In line with the suggestions of the WHO (2010, 2012), Smith Slep et al. (2014) stated that “ecological models are especially compatible with a public health perspective, which emphasizes a need to focus on modifiable factors and view individuals as embedded within nested systems that can be leveraged to affect change” (p. 471). Thus, in addition to explaining why IPV occurs, the ecological model can also be used to provide an understanding of how such violence can be prevented (e.g., Heise, 1998; WHO, 2010).

In contrast to previous single-factor theories relying on either psychological or feminist ideological and political explanations, the ecological model assumes that the etiology of IPV is more complex (Heise, 1998). Hence, the model assumes that the correlates of IPV best can be explained as the result of several different risk factors operating on four different levels: individual, family, community, and society (see Figure 1). Importantly, to understand the complex nature of IPV, researchers have emphasized the need to understand the interaction among risk factors on different levels in the model (Heise, 1998; Lauritsen & Schaum. 2004). This is illustrated in the model in that the different levels are nested within each other (e.g., individual risk factors influence the perpetrator’s interaction with the family level, as well as with the community and society levels).

(17)

7

Figure 1. The integrated ecological model of intimate partner violence (inspired by Heise, Ellsberg, & Gottemueller, 1999).

The four different levels each contain different risk factors for IPV (Heise, 1998; WHO, 2012). The individual level mainly consists of factors related to developmental experiences and personal characteristics that each individual brings with them into an intimate relationship. For example, such risk factors include being abused during childhood, personal attitudes and beliefs, impulsivity, or mental health problems. The family level concerns relational and situational risk factors for IPV. Situational risk factors relate to the perpetrator's interaction with others and the settings where such interactions take place. In the case of IPV, this refers to the interaction between the perpetrator and the female partner, and the setting consists of the family/relationship. For example, such risk factors could include intimate relationship problems in the form of a perpetrator's inadequate relationship skills, which could entail resorting to violence to solve conflicts with the partner. Furthermore, researchers argue that alcohol and drug abuse are situational risk factors for IPV as these, among other things, reduce inhibitions, and impair the individual’s ability to interpret social cues (Heise, 1998). Risk factors on the community level operate on a wider situational level in which the family level is embedded within. As such, the community level refers to both formal and informal risk factors for IPV. Risk factors for IPV related to the former include employment problems and low socioeconomic status, whereas the latter refers to risk factors such as isolating a partner socially through controlling behaviors. Finally, the societal level in the model contains risk factors for IPV on a structural level, such as attitudes, values, and beliefs that condone violence against women or support male entitlement of women.

(18)

8

However, according to the WHO (2012), the influence of societal-level risk factors on IPV perpetration is more ambiguous than risk factors on the individual and family levels. As such, there is limited research of risk factors on the societal level, and such risk factors have also been shown to be context specific (e.g., varying both between and within countries).

To illustrate the use of the ecological model, consider a man with low self-control (individual level) who may not ordinarily use violence against his partner. However, losing his job (community level), rendering his female partner the main provider of the household, may create marital conflict (family level) related to financial support issues, which could escalate into the use of physical violence. The use of IPV in this scenario might be even more likely if these events occur within a culture where masculinity is defined in terms of being the main provider for the family (societal level). This brief scenario combines risk factors on all four levels within the ecological model to explain the violence etiology for this perpetrator. Importantly, this scenario does not aim to justify the use of IPV nor minimize the perpetrator’s personal responsibility for the violent act. Rather, this illustration is used to provide a possible etiology of IPV for a certain type of partner violent man.

As concluded by Heise (1998), the ecological model can be used to understand differences among IPV perpetrators. As such, it can provide a framework for understanding why some IPV perpetrators are violent in certain situations, whereas others are not (Heise, 1998; Holtzworth-Munroe & Stuart, 1994; Simmons & Lehmann, 2010). To this end, mapping the risk factors for the IPV perpetrators onto the ecological model could help better explain and understand why they use IPV. Such a mapping procedure could also assist in the planning of risk management strategies (Heise, 1998; Smith Slep et al., 2014). This could facilitate decisions by those who respond to IPV perpetrators, such as the police, in terms of which levels within the ecological model that interventions need to address to prevent recidivism for a specific perpetrator.

1.4.2 Stake in conformity

In line with the focus of the criminal justice approach on punishment and deterrence of IPV perpetrators (WHO, 2010), the criminological theory of stake in conformity attempts to explain why some individuals refrain from criminal and antisocial behavior. It was therefore used as a complement to the ecological model to understand and explain the use of IPV among partner violent men.

(19)

9

The stake in conformity theory was originally proposed by Jackson Toby (1957) and builds on the criminological theory of social disorganization (Shaw & McKay, 1942), which states that crime occurs because of diminished social vigilance. Social disorganization theory belongs to the branch of criminological theories known as control theories. The common denominator of such theories is their attempt to explain why the majority of individuals refrain from a criminal lifestyle. In this sense, Toby (1957) argued that social disorganization in itself was limited in explaining the causes of crime, mainly because the theory failed to acknowledge individual differences: not everyone who is brought up under weak parental and community control becomes a criminal. More specifically, Toby (1957) claimed that these individual differences could be accounted for by examining to what extent an individual could be seen as having a stake in the pro-social (i.e., conform) values advocated by society. Such values were operationalized as educational, relational, and occupational success.

Toby (1957) argued that, compared to individuals who have weak ties to conventional society, individuals who have strong ties (i.e., being married, employed and having a high level of education) have more to lose in committing a crime. Such individuals not only risk losing their job and their marriage, but perhaps more importantly, they risk the loss of social approval and social status among friends and neighbors (Goldstein, Cantos, Brenner, Verborg, & Kosson, 2016; Klein, Wilson, Crowe, & DeMichele, 2008; Toby, 1957). On the other hand, individuals with weaker ties to conventional society (i.e., unmarried, unemployed, and low levels of education) already consider themselves as outsiders and have less to lose in terms of the social reactions of others to their criminal behavior (Toby, 1957). Thus, it was hypothesized that such individuals were more inclined to commit crimes.

The stake in conformity theory can be used to discuss and analyze the resulting risk profiles and recidivism rates of partner violent men in this thesis. To the extent that this thesis can demonstrate different characteristics attributed to pro-social and antisocial traits, as well as recidivism rates, this theory can provide a valuable theoretical framework for discussing and explaining such findings. Importantly, there is no commonly accepted definition of stake in conformity (Mach, Cantos, Weber, & Kosson, 2017), which means that the operationalization of this theory is not limited to measures of educational, relational, and occupational attainment.

Finally, the stake in conformity theory could also be applied in combination with the integrated ecological model. For instance, IPV perpetrators who have a high stake in conformity should have few, if any, risk

(20)

10

factors for violence on the community level. According to the theory, perpetrators with a high stake in conformity want to avoid the potential risk of losing their social status and social position (Toby, 1957). As such, their main risk factors should be on the individual and family levels within the model. Thus, the results generated by this thesis will also be discussed in this integrated fashion, combining the ecological model with the stake in conformity theory.

1.5 Policing IPV

As previously mentioned, the police are one of the primary responders to IPV. Although IPV has always existed, it was only quite recently that this form of violence was defined as a crime, thus, justifying interventions by the criminal justice system (Erez, 2002; European Union Agency for Fundamental Rights, 2014). Traditionally, IPV has been considered as a private matter where police would intervene only when the victim suffered a serious injury (Erez, 2002). However, in the 1980s, an increasing demand on law enforcement to solve social issues combined with the political influence of the women's movement, police responses to IPV became more common. This response was mainly reactive in terms of arresting the alleged perpetrator, in part due to the results of the Minneapolis Domestic Violence Experiment (Sherman & Berk, 1984). This social experiment found that arresting the alleged perpetrator was the most efficient method (compared to mediation and temporarily separating the parties) in preventing recidivism. As such, arrested perpetrators were less likely to recidivate in IPV than those perpetrators who were not arrested. These results were pivotal in the introduction of mandatory arrest laws for IPV perpetrators in the United States (Feder, 1999). However, subsequent attempts to replicate the seemingly deterrent effect of arrest on recidivism have shown mixed results. Analyzing data from all five replication attempts of the Minneapolis Domestic Violence Experiment, results showed that arrest had a modest deterrent effect when measuring recidivism via victim reports, and no deterrent effect when measuring recidivism using official police records (Maxwell, Garner, & Fagan, 2001).

Albeit, the traditionally reactive stance characterizing police responses to IPV has shifted over the last decades. Nowadays, there is a stronger focus on preventing this form of violence (Campbell, Gill, & Ballucci, 2018). To this end, a widely used prevention strategy adopted by police internationally is the use of violence risk assessments (e.g., Campbell et al., 2018; Kropp, 2008). Violence

(21)

11

risk assessment is a method for criminal justice professionals to manage violent situations, such as police officers responding to IPV-related calls (Kropp, 2008). In the words of Trujillo and Ross, “assessing and responding to risk are key elements in how police respond to domestic violence” (2008, p. 454).

1.5.1 Violence risk assessments

Internationally, the police use different approaches to assess and respond to risk for future IPV (Nicholls, Pritchard, Reeves, & Hilterman, 2013). To understand risk assessment one must first define the term 'risk'. In its most straightforward sense, risk is the likelihood that an individual will engage in a certain type of behavior (Heilbrun, Yashara, & Shah, 2010). However, according to Kropp (2008), the risk for violence should be understood more broadly than just the likelihood that an individual will commit a violent act in the future. Additionally, such an understanding of risk must also include considerations of the imminence, nature, frequency, and severity of such potential violence.

In assessing the risk for future violence, the assessor either implicitly or explicitly makes use of risk factors. The implicit use of risk factors refers to cases where no guidelines or checklists are used to identify such factors, whereas the explicit use of risk factors refers to cases where guidelines or checklists are used. In turn, risk factors can be defined as historical, demographical, biological, cultural, or situational characteristics, which are related to violent behavior (Jackson, 1997, cited in Webster, & Hucker, 2007). Thus, in a risk assessment context, risk factors refer to factors that increase the risk for violence (e.g., Heilbrun et al., 2010; Kropp, 2008; Webster & Hucker, 2007). Furthermore, risk factors can be either static or dynamic (Heilbrun, 1997). The former refers to historical factors that cannot be changed through intervention (e.g., sex and prior convictions), whereas the latter refers to factors that can be altered by interventions (e.g., substance abuse and mental health problems).

Overall, when it comes to assessing the risk for future violence there are mainly three approaches: the unstructured clinical approach, the actuarial approach, and the structured professional judgment (SPJ) approach (Douglas & Kropp, 2002; Kropp, 2008). The unstructured clinical approach is considered the most widely used approach for assessing risk for future IPV (Kropp, 2008). Here, the risk assessor estimates the risk for future violence relying solely on his or her professional experience, knowledge, and intuition of risk. This approach is unstructured, as the assessor does not rely on any

(22)

12

structured guidelines (e.g., checklists or tools). As a result, this approach to violence risk assessment has been criticized for lacking validity and reliability (Kropp, 2008), as well as being too subjective and informal (Grove & Meehl, 1996).

The actuarial approach to risk assessment consists of making formal predictions of risk. As such, this approach relies on calculating probabilities (e.g., using equations or formulas) for the risk of future violence (Heilbrun et al., 2010), usually within a specific time-period (Douglas & Kropp, 2002). To calculate such a probability, the actuarial approach to risk assessment makes use of structured risk assessment instruments containing risk factors. Such risk factors are empirically derived from specific samples of offenders and included in the instruments due to their predictive validity (i.e., their positive association with increased risk for violence: Heilbrun et al., 2010). Therefore, the actuarial approach to risk assessment is a significant improvement compared to the unstructured clinical approach in terms of reliability and validity (e.g., Ægosdottír et al., 2006; Heilbrun et al., 2010). Although the actuarial approach facilitates the assessment of the level of risk for repeat violence, it has several shortcomings. Most importantly, it does not usually provide the assessor with any guidelines for how to minimize or prevent the predicted risk (e.g., Douglas & Kropp, 2002; Hart, 1998; Heilbrun, 1997). As such, the primary aim of violence risk assessments should be to prevent violence, rather than merely predicting it (e.g., Hart, 1998).

To this end, the SPJ approach attempts to combine the structure of the actuarial approach (using structured risk assessment instruments containing evidence-based risk factors) with the flexibility of the unstructured clinical approach to include the professional's discretion and expertise (Douglas & Kropp, 2002). To achieve this, SPJ risk assessment instruments contain risk factors for violence, which have been derived based on broad systematic literature reviews (Heilbrun et al., 2010). However, the included risk factors in specific SPJ instruments are not exclusive; other important, case-specific, risk factors are allowed for consideration (Hart, 1998). This addresses an issue with actuarial instruments where professionals (e.g., police officers) feel that their expertise and discretion are overlooked (Hart, 1998). As such, unlike the actuarial method, the professional's expertise and discretion should be the guiding principle when making final decisions about the relative, or combined, contribution of a perpetrator's risk factors to the overall risk for violence (Douglas & Kropp, 2002). When such a summary risk rating for violence has been made, the assessor needs to suggest relevant risk management strategies aiming to reduce or prevent the estimated risk. As SPJ

(23)

13

instruments place a greater focus on dynamic risk factors, risk management strategies should focus on addressing such factors that can be changed through interventions (Hart, 1998).

As mentioned above, the prevention of violence should be the ultimate goal of risk assessment (Douglas & Kropp, 2002; Hart, 1998). As such, risk assessments should help guide the choice of relevant risk management strategies. Overall, risk management strategies can be organized into four kinds of activities: monitoring, treatment, supervision, and victim safety planning (Kropp, 2008). Monitoring refers to evaluating changes in risk for violence over time so that risk management strategies can be revised accordingly. However, unlike supervision or control, monitoring is less intrusive for the perpetrator and the victim, and could constitute of telephone meetings, home visits, or contact with service care providers (e.g., health care, social services, and therapists) for updates on, for example, missed appointments. Treatment refers to a broad variety of interventions aiming to reduce the risk of violence. This can include treatment interventions for mental health problems, substance abuse, violent attitudes, and relationship functioning. Supervision entails, for instance, incapacitation and community supervision. Such forms of supervision aim to make it more difficult for the perpetrator to recidivate in violence. Finally, victim safety planning aims to improve the victim's dynamic and static resources. The former relates to the victim's social resources (e.g., therapy or training in self-protection), whereas the latter relates to the victim's physical environment (e.g., providing the victim with personal alarms).

1.5.2 Police risk assessments of IPV

Although Kropp (2008) noted that the clinical unstructured approach is probably the most widely used when it comes to IPV risk assessments, there are several actuarial and SPJ-based IPV risk assessments instruments available to, and used by, the police (e.g., Kropp, 2008; Nicholls et al., 2013). This includes the Ontario Domestic Assault Risk Assessment (ODARA: Hilton, Harris, Rice, Lang, & Cormier, 2004), the Danger Assessment (DA: Campbell, 1995), the Spousal Assault Risk Assessment guide (SARA: Kropp, Hart, Webster, & Eaves, 1999), and the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER: Kropp et al., 2010). Being actuarial or SPJ-based, these risk assessment instruments have all been developed on a scientific basis, thus containing evidence-based risk factors for IPV.

Of these instruments, the SARA (Kropp et al., 1999) is a commonly used SPJ instrument within the police internationally. As such, it has been

(24)

14

translated into at least 10 languages and used in at least 15 countries (Kropp & Gibas, 2010). Importantly, the SARA instrument employs a wider definition of IPV than its actuarial counterparts do, where the ODARA cannot be used with dating or same-sex relationships and the DA focus on lethal violence. Although the SARA has been used within the police for a long time, the authors of this instrument realized that it was not optimal for non-clinically trained professionals, such as the police (Kropp et al., 2010). For instance, the SARA consists of 20 risk factors measuring criminal history, psychosocial adjustment, spousal assault history, index offense, and other considerations, which is both time-consuming to assess and require specific judgments of mental health problems. Thus, a shorter version of the SARA, the B-SAFER, was created for police officers to use. The B-SAFER was reduced to ten risk factors related to the perpetrator, as well as included five victim vulnerability factors. Subsequently, the B-SAFER was developed and pilot-tested by police officers in both Canada and Sweden (Kropp et al., 2010). Today, the Swedish version of the B-SAFER (Kropp, Hart, & Belfrage, 2008) is prescribed by the NPB (2017) to be used in all seven police regions in Sweden.

The B-SAFER. The B-SAFER (Kropp et al., 2008, 2010) contains five risk factors related to the perpetrator’s history of IPV, five risk factors related to the perpetrator’s psychosocial adjustment, and five victim vulnerability factors (see Table 1). In terms of the first five risk factors, violent acts refer to the actual or attempted use of physical violence, including sexual violence and the use of weapons. Violent threats or thoughts include oral or written threats of physical harm that are communicated directly or indirectly to the victim, as well as plans, fantasies, and urges to harm the victim. Threats also include following the victim (stalking), using threatening non-verbal communication, and the brandishing of weapons. Escalation means that the previous two risk factors (i.e., violent acts, and violent threats or thoughts) have increased in severity, frequency, or diversity over time. Violation of court orders includes breaching conditions of bail, parole, probation, and restraining orders. Violent attitudes refer to attitudes, beliefs, values or thoughts that support, condone, or minimize the use of IPV. This includes socio-political, religious, cultural, sub-cultural and personal beliefs that excuse or justify violent and controlling behavior against a partner. This is often expressed as sexual jealousy, patriarchal beliefs, denial of violent behavior, and entitlement of women.

According to the B-SAFER manual (Kropp et al., 2010), the psychosocial adjustment risk factors refer to the psychological and social functioning of the perpetrator. To this end, general criminality includes tendencies to engage in

(25)

15

persistent, frequent, or diverse antisocial behavior, such as criminal conduct and general violence. This risk factor also entails endorsement of general antisocial attitudes (e.g., condoning or justifying behaviors that violate others) and associating with antisocial peers. Intimate relationship problems reflect difficulties of establishing and maintaining stable romantic relationships, free from conflicts such as frequent break-ups, divorce, or infidelity. Employment problems refer to a perpetrator's problems related to seeking or maintaining gainful employment, frequent job changes, poor work performances, and the serious financial problems associated with long periods of unemployment. Substance use problems include using illegal drugs and the misuse of legal drugs (e.g., alcohol), to the extent that it results in substantial impairment of the individual’s health or social functioning (e.g., overdose, job loss, and arrest). Finally, mental health problems include impairment of mental health, for instance, in the form of psychosis (e.g., delusions, and hallucinations), personality disorders (e.g., depression and anxiety), or extreme anger or impulsivity.

Table 1. Risk and victim vulnerability factors in the B-SAFER (Kropp, Hart, & Belfrage, 2010)

Perpetrator risk

factors Psychosocial adjustment Victim vulnerability factors Violent acts General criminality Inconsistent behavior/attitude Violent threats or

thoughts Intimate relationship problems Extreme fear

Escalation Employment problems Inadequate access to resources Violation of court

orders Substance use problems Unsafe living situation

Violent attitudes Mental health problems Personal problems Note. B-SAFER = Brief Spousal Assault Form for the Evaluation of Risk.

The victim vulnerability factors refer to factors that may affect the victim’s propensity to cooperate in the risk management strategies and thus facilitate victim protection (Kropp et al., 2010). These factors pertain to the victim’s behavioral (e.g., minimizing the perpetrator’s violence), psychosocial (e.g., substance abuse or unemployment), and situational vulnerability (e.g., being socially isolated, or having insufficient physical or dynamic security for protection against the perpetrator), which directly or indirectly increase the risk of being victimized by IPV. The victim vulnerability factors are in no way

(26)

16

meant to shift the blame or guilt for being victimized to the victim. Rather, by assessing these factors the likelihood that the victim will participate in any recommended risk management can be more accurately determined (Kropp, Hart, & Lyon, 2006).

In line with the SPJ approach to violence risk assessment, other risk and victim vulnerability factors can also be considered and included in the risk assessment (Kropp et al., 2010). For instance, this could include a perpetrator’s access to guns, or having children together with the perpetrator. Other case-specific factors can also be relevant, as long as they likely either play a causal role for a perpetrator’s use of violence, or a victim’s ability to take self-protective actions.

The risk and victim vulnerability factors described above are rated on a three-point scale as Present (Yes), Partially present (To some extent) or Absent (No). Information regarding the risk and victim vulnerability factors should be acquired from multiple sources, preferably from hearings with the victim, the alleged perpetrator, and potential witnesses, as well as the police crime registers. Ideally, the risk assessment should be based on all these information sources. In practice, however, not all sources are available to the assessor. It is important, however, that these risk assessments are victim-informed (Kropp, 2008). As mentioned above, the alleged perpetrators often underreport their use of violence, whereas the victim's self-report is a more reliable source of information (even if victims, of course, also for various reasons can choose to withhold important information: Kropp, 2008). Lack of information concerning a specific risk or victim vulnerability factor may result in its exclusion.

The presence or absence of the perpetrator risk factors and psychosocial adjustment are considered in the current situation (i.e., referring to the current crime, or crimes, for which the perpetrator have been subjected to a risk assessment) and in the past (i.e., at any point in time before the current situation). Ultimately, drawing on the presence and relevance of the B-SAFER risk and vulnerability factors for a specific IPV case, the assessor produces two summary risk ratings: one pertaining to the risk for acute IPV (i.e., the risk for imminent violence), and one pertaining to the risk for severe or deadly IPV. These summary risk ratings are rated as low, moderate, or high risk for future violence.

The B-SAFER has proven to be a useful tool for both the Swedish police (e.g., Belfrage & Strand, 2008), as well as for police internationally (Kropp et al., 2010). The instrument is reported to have good-to-excellent interrater reliability and predictive validity (e.g., Gerbrandij, Rosenfeld, Nijdam-Jones,

(27)

17

& Galietta, 2018; Loinaz, 2014; Serie, van Tilburg, van Dam, & de Ruiter, 2017; Storey et al., 2014; Thijssen & de Ruiter, 2011), as well as shown good concurrent and construct validity (Au et al., 2008). Finally, the B-SAFER has shown valid and reliable results for the Swedish police. For instance, Belfrage and Strand (2008) found a strong positive correlation between the number of B-SAFER risk factors rated as present and the police officers’ assessed degree of risk. As such, a greater number of risk factors rated as present were related to higher summary risk ratings. Moreover, Storey et al. (2014) reported that the Swedish police were capable of performing B-SAFER assessments that were related to risk management, where higher summary risk ratings resulted in more risk management strategies proposed.

1.5.3 Policing IPV in Sweden

In January 2015, the Swedish police underwent a major re-organization. Initially consisting of 21 police authorities, the Swedish police were centralized into one authority consisting of seven geographical regions (The Swedish Police, 2020). Due to this centralization, the Swedish police regions work uniformly according to the national guidelines for policing IPV (NPB, 2019). This includes responding to IPV-related emergency calls, assisting the prosecutor in carrying out the criminal investigation (e.g., conducting hearings with the victim and the suspect), and being responsible for the victim's safety by attempting to prevent the risk for recidivism.

Importantly, the work with IPV cases is organizationally divided into two units: a criminal investigation unit and a crime victim support unit (NPB, 2017). These two units work separately with the same cases, but with different priorities. As such, the investigative work should be objective in terms of collecting evidence from both parties (i.e., the suspect and the alleged victim) to determine if a crime has been committed and whether or not there are legal grounds to prosecute the suspect. Parallel to the criminal investigation, the victim support unit works exclusively with the victim. This includes providing support and services necessary for the victim's safety and wellbeing. Regardless of whether the case proceeds to court or not, the suspect may still pose a threat to the victim. In such cases, the victim may require risk management strategies in the form of protective actions (e.g., protected living or alarm devices) to minimize the risk of being repeatedly victimized. To determine the perpetrator's risk for recidivism and the victim's need for protective actions, the current national police guidelines state that the Swedish police must rely on violence risk assessments (NPB, 2019).

(28)

18

These violence risk assessments are carried out on different levels in different stages of the process of a police-reported IPV crime. To begin with, whenever an IPV-related crime is reported, Swedish police officers are obliged to make an initial screening assessment of the imminent risk for violence or threats directed against the victim (NPB, 2019). If such a risk is perceived as evident, the case is referred to the crime victim support unit, which is responsible for carrying out structured violence risk assessments using the B-SAFER (Kropp et al., 2008) as well as to suggest adequate risk management strategies.

In terms of risk management strategies, the police have access to several victim-oriented protective actions. These protective actions can be divided into dialogue (e.g., communicating the outcome of the risk assessments via safety talks), technical protection (e.g., alarm devices), administrative actions (e.g., providing the victim with fictitious personal information), surveillance, or protected living (NPB, 2012). Except for some of the administrative protective actions, the victim-oriented protective actions require the victim’s consent and co-operation. Meanwhile, there are fewer available actions directed towards the suspect aiming to minimize his risk for recidivism. The suspect-oriented actions available to the police consist of restraining orders, surveillance, and proactive talks. However, in Sweden, the decision to grant a restraining order is made by the prosecutor and not the police. The purpose of proactive talks is to make the perpetrator understand the negative consequences of using violence in an intimate relationship (NCCP, 2017). However, the proactive talks are only carried out if the police consider the perpetrator susceptible to such a talk. This excludes perpetrators with, for example, mental health problems and those perpetrators who lack insight into their violent behavior. Problematically, this limits the number of perpetrators eligible for this form of protective action.

In terms of its crime preventive effect, the only suspect-oriented risk management strategy that has been evaluated in Sweden is the use of restraining orders (Strand, 2012). Results from this evaluation demonstrated that the restraining orders had a limited overall effect in preventing IPV recidivism. However, the results indicated that the restraining orders had a crime preventive effect for those perpetrators who were assessed with the B-SAFER as having a lower risk for IPV recidivism. Thus, these results are in line with previous conclusions stating that the crime preventive effect of legal interventions most likely depends on the perpetrator type (e.g., Sherman & Berk, 1984; Williams, 2005).

(29)

19

1.6 IPV typologies

The previous sections have highlighted the diversity of IPV in terms of various risk factors for violence, variations in recidivism rates, and the different effect of legal interventions on recidivism for perpetrators of such violence. Thus, it becomes evident that IPV is a complex and heterogeneous form of violence against women. A commonly used procedure to understand this heterogeneity is to classify both IPV and perpetrators of such violence into typologies (Ali, Dhingra, & McGarry, 2016). Typologies can be defined as “a means of classifying or categorizing subject matter into groups” (Boxall, Rosevear, & Payne, 2015, p. 1). Such groups are often labeled as “subtypes”. Thus, a typology consists of several distinct subtypes, which differ on the variable (or variables) of interest.

In the broadest sense, attempts have been made to classify the type of IPV into subtypes (see Ali et al., 2016 for a review). To this end, the most influential typology was proposed by the American sociologist Michael Johnson (1995) who initially suggested a typology of IPV consisting of patriarchal terrorism and common couple violence (later referred to as intimate terrorism and situational couple violence, respectively: Johnson & Leone, 2005). This typology was created against the backdrop of the two competing theoretical and policy views of the family violence scholars and the feminist scholars. As such, relying on large-sample national survey methodology, the family violence perspective found that IPV was often characterized by bidirectional violence and usually the result of occasional marital conflicts leading to minor forms of violence (i.e., situational couple violence). In contrast, relying mainly on qualitative interviews with women drawn from hospitals and women's shelters, the feminist perspective emphasized the role of structural causes of IPV (e.g., patriarchal beliefs, and gender inequality) that rendered a power imbalance between genders in men's favor (i.e., intimate terrorism). As such, Johnson argued that family violence scholars and feminist scholars, due to their differing methodologies, study two different phenomena (Johnson, 1995; Johnson & Leone, 2005). Importantly, unlike the common misconception, the severity and frequency of IPV do not differ between situational couple violence and intimate terrorism (Johnson & Leone, 2005). Instead, the main distinction between these two forms of IPV is the level of control. Accordingly, “intimate terrorism is violence that is embedded in a general pattern of control; situational couple violence is not” (Johnson & Leone, 2005, p. 324).

However, in terms of IPV typology research, the focus has been on identifying subtypes of male perpetrators of such violence. According to

(30)

20

Simmons and Lehmann (2010), the purpose of such research is to identify distinct subtypes so that interventions can be tailored according to a perpetrator’s specific risk profile. To this end, the most influential perpetrator typology to date was suggested by Holtzworth-Munroe and Stuart (1994). In their seminal paper, acknowledging the heterogeneity among IPV perpetrators, they argued that the development of a male IPV perpetrator typology would shed light on how and why men use violence against their partner. In turn, such knowledge could guide the tailoring of interventions according to a perpetrator’s risk profile so that IPV is more efficiently reduced. As such, Holtzworth-Munroe and Stuart (1994) reviewed 15 studies, both deductive (i.e., studies creating subtypes on a priori relevant variables and then comparing them on available data) and inductive (i.e., studies using cluster analytic approaches to create subtypes), of IPV perpetrator typologies. Their review aimed to identify the subtypes that most consistently appeared across studies. Based on the results of their review, the authors proposed the existence of three subtypes of male IPV perpetrators: the family-only perpetrator, the dysphoric/borderline perpetrator, and the generally violent/antisocial perpetrator.

Holtzworth-Munroe and Stuart (1994) suggested that these subtypes could be classified on a three-level scale of low, moderate, or high, on the following three dimensions: a) the severity and frequency of violence, b) the generality of violence (including criminal history and legal involvement), and c) psychopathology or personality disorders. The family-only perpetrator was described as using low levels of non-severe violence on an infrequent basis, being violent only within the family, as well as having little psychopathology and either no personality disorder or possibly a passive-dependent personality disorder. The dysphoric/borderline perpetrator was suggested to use moderate to high levels of violence (including more severe forms of physical, psychological and sexual abuse), being violent primarily towards the partner but also occasionally violent towards others (e.g., strangers or friends), and displayed traits of borderline personality disorder and severe depression (i.e., dysphoria). The generally violent/antisocial perpetrators (hereafter referred to as antisocial for brevity) were hypothesized to use severe forms of violence frequently (including psychological and sexual violence) towards their partner, being violent also toward others, and to display traits of antisocial personality disorder and psychopathy.

Finally, Holtzworth-Munroe and Stuart (1994) concluded by encouraging researchers to validate their proposed, theoretically based, typology of IPV perpetrators. However, such validation attempts have produced mixed

(31)

21

results. First, studies differ in the number of subtypes identified, where typology solutions ranging between two and five subtypes have been suggested as most appropriate (e.g., Chiffriller & Hennessy, 2006; Graña, Redondo, Muñoz-Rivas, & Cantos, 2014; Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart, 2000; Huss & Ralston, 2008; Serie et al., 2017; Tweed & Dutton, 1998; Walsh et al., 2010). Second, although some studies have found support for the three-subtype typology proposed by Holtzworth-Munroe and Stuart (1994), problems have been reported for both researchers and clinicians to distinguish between the subtypes. For instance, several studies have reported difficulties distinguishing the dysphoric/borderline perpetrators from antisocial perpetrators on key features such as traits of borderline personality disorder (e.g., Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart, 2003; Huss & Ralston, 2008; Stoops, Bennett, & Vincent, 2010). Additionally, difficulties among clinicians to accurately categorize IPV perpetrators into the three pre-existing subtypes have been reported (Langhinrichsen-Rohling, Huss, & Ramsey, 2000; Lohr, Bonge, Witte, Hamberger, & Langhinrichsen-Rohling, 2005). Relatedly, the studies originally reviewed by Holtzworth-Munroe and Stuart (1994), as well as subsequent efforts to validate their suggested typology, have almost exclusively relied on clinical instruments and advanced statistical techniques (e.g., cluster analytic approaches) to identify the subtypes. This has raised concerns over the practical utility of such typologies for non-clinically trained criminal justice professionals (e.g., police officers and probation officers) working with IPV perpetrators (Boxall et al., 2015; Cantos, Goldstein, Brenner, O’Leary, & Verborg, 2015; Stoops et al., 2010). Criminal justice professionals rarely have the means, in terms of resources, information, or time to undertake the level of analysis necessary to identify subtypes based on such complex methods (Boxall et al., 2015; Lohr et al., 2005).

In sum, the proposition of the three-subtype typology offered by Holtzworth-Munroe and Stuart (1994) has had a huge impact on the perception of IPV perpetrators. Shifting from an erroneous homogeneous view of such perpetrators into an accurate heterogeneous view, the prospect of preventing IPV through the tailoring of interventions according to the specific risk profiles of perpetrator subtypes has greatly increased (e.g., Cantos & O’Leary, 2014). However, as presented above, there are several important limitations with Holtzworth-Munroe and Stuart’s (1994) typology. If IPV perpetrator typologies are to fulfill their potential usefulness in facilitating the prevention of IPV recidivism (e.g., Cavanaugh & Gelles, 2005; Holtzworth-Munroe & Stuart, 1994), they have to be both valid as well as

(32)

22

possible to use for criminal justice professionals such as the police. Moreover, Holtzworth-Munroe and Stuart (1994) concluded that in order to generate a practically useful typology, “researchers should identify the smallest subsets of measures needed to reliably distinguish between batterer subtypes” (p. 494).

1.7 A novel approach to subtyping IPV perpetrators

Bearing the abovementioned limitations in mind, a handful of studies have adopted an easier and more straightforward approach to subtyping IPV perpetrators. This approach categorizes such perpetrators based on the generality of violence dimension (as suggested by Holtzworth-Munroe & Stuart, 1994) as either antisocial or family-only. Thus, this novel approach relies on a taxonomical rather than dimensional classification procedure to create subtypes of partner violent men (cf. Holtzworth-Munroe & Stuart, 1994). Although it can be debated whether a taxonomical or dimensional approach to categorizing subtypes are most accurate, Holtzworth-Munroe and Meehan (2004) concluded that both approaches usually generate similar results. As such, studies adopting this taxonomical procedure has successfully managed to identify subtypes resembling the antisocial and the family-only subtypes (Boyle, O’Leary, Rosenbaum, & Hassett-Walker, 2008; Cantos et al., 2015; Wang, Horne, Holdford, & Henning, 2008). To this end, antisocial perpetrators are defined as those perpetrators who are violent towards their partner but also towards others outside of the relationship (i.e., generally violent), whereas family-only perpetrators are defined as perpetrators who are violent only towards their partner (i.e., non-generally violent).

Moreover, the antisocial and the family-only subtypes resulting from this novel approach have been found to differ in a manner that is theoretically consistent with Holtzworth-Munroe and Stuart’s (1994) original description of these two subtypes on all three dimensions (i.e., severity and frequency of violence, generality of violence, and psychopathology). As such, these subtypes have been found to differ on several behavioral and psychosocial traits. In terms of behavioral traits, antisocial perpetrators have demonstrated a more extensive criminal history as well as an earlier onset of antisocial behavior compared to family-only perpetrators (Boyle et al., 2008; Cantos et al., 2015). As such, in the study by Boyle et al. (2008), relying on self-reports on various personality and criminal history scales, antisocial perpetrators reported engaging in more lifetime general antisocial and violent behavior than family-only perpetrators. Antisocial perpetrators also displayed more

(33)

23

behavioral disinhibition than the family-only perpetrators, in terms of conduct disorder and delinquency, as well as being more psychologically abusive towards their partner (Boyle et al., 2008). Relatedly, Wang et al. (2008) found that the pattern of IPV was different for the two subtypes. To this end, family-only perpetrators' use of IPV was more highly correlated to the victim's use of violence compared to for the antisocial perpetrators. This seems to indicate that the family-only perpetrators are more likely to be involved in bidirectional violence compared to the antisocial perpetrators.

Differences between the antisocial and the family-only perpetrators have also been found concerning psychosocial risk factors. As such, the antisocial perpetrators have evidenced more psychosocial risk factors than family-only perpetrators. This includes more problems with substance use (both alcohol and drugs) and employment (Cantos et al, 2015; Wang et al., 2008), as well as displaying more mental health problems (Boyle et al., 2008; Cantos et al., 2015). In terms of mental health problems, such findings demonstrate that the antisocial perpetrators are more likely to display traits related to Cluster B personality disorders (e.g., antisocial personality and borderline disorder) and psychopathy (Boyle et al., 2008).

However, although studies have found several differences between the antisocial and the family-only perpetrators, two major limitations surround this body of knowledge. First, only a few studies have used this novel approach to study differences between these perpetrators, and such studies have examined clinical (Boyle et al., 2008; Wang et al., 2008) and probation samples (Cantos et al., 2015). Thus, it is uncertain whether the behavioral and psychosocial differences between the antisocial and the family-only perpetrators would apply to a sample of police-reported IPV perpetrators. As such, knowledge of these subtypes’ risk profiles could prove useful for the police in their attempts to identify and prevent the high-risk perpetrators from recidivating. Second, although Cantos et al. (2015) reported that antisocial perpetrators were assessed with a higher risk for IPV recidivism compared to the family-only perpetrators, none of the studies described above has examined actual recidivism rates among these subtypes. As previously mentioned, the potential usefulness of IPV perpetrator typologies lies in their ability to help guide decisions related to the subtypes' level of risk for recidivism and what type of risk management interventions that are most fitting to reduce this risk (e.g., Cavanaugh & Gelles, 2005). Obviously, such decisions require knowledge about the potential differences in risk for recidivism between the antisocial and the family-only subtypes.

Identifying risk for recidivism among partner violent men reported to the Swedish police